Brain tumors are among the most fatal and devastating diseases, often resulting in significantly reduced life expectancy. An accurate diagnosis of brain tumors is crucial to devise treatment plans that can extend the lives of affected individuals. Manually identifying and analyzing large volumes of MRI data is both challenging and time-consuming. Consequently, there is a pressing need for a reliable deep learning (DL) model to accurately diagnose brain tumors. In this study, we propose a novel DL approach based on transfer learning to effectively classify brain tumors. Our novel method incorporates extensive pre-processing, transfer learning architecture reconstruction, and fine-tuning. We employ several transfer learning algorithms, including Xception, ResNet50V2, InceptionResNetV2, and DenseNet201. Our experiments used the Figshare MRI brain tumor dataset, comprising 3,064 images, and achieved accuracy scores of 99.40%, 99.68%, 99.36%, and 98.72% for Xception, ResNet50V2, InceptionResNetV2, and DenseNet201, respectively. Our findings reveal that ResNet50V2 achieves the highest accuracy rate of 99.68% on the Figshare MRI brain tumor dataset, outperforming existing models. Therefore, our proposed model's ability to accurately classify brain tumors in a short timeframe can aid neurologists and clinicians in making prompt and precise diagnostic decisions for brain tumor patients.

翻译：脑肿瘤是最致命且最具破坏性的疾病之一，常导致患者预期寿命显著缩短。准确诊断脑肿瘤对于制定可延长患者生命的治疗方案至关重要。手动识别和分析大量MRI数据既困难又耗时。因此，迫切需要一种可靠的深度学习模型来准确诊断脑肿瘤。本研究提出了一种基于迁移学习的新型深度学习方法，用于有效分类脑肿瘤。我们的新方法整合了广泛的预处理、迁移学习架构重构与微调。我们应用了多种迁移学习算法，包括Xception、ResNet50V2、InceptionResNetV2和DenseNet201。实验采用Figshare MRI脑肿瘤数据集（包含3064张图像），Xception、ResNet50V2、InceptionResNetV2和DenseNet201分别取得了99.40%、99.68%、99.36%和98.72%的准确率。研究结果表明，ResNet50V2在Figshare MRI脑肿瘤数据集上实现了99.68%的最高准确率，优于现有模型。因此，本模型能在短时间内准确分类脑肿瘤的能力，可辅助神经科医生与临床医师为脑肿瘤患者做出及时精准的诊断决策。